ModelSociology.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import findspark\n",
    "findspark.init('/home/sdmohant/spark-3.0.0-bin-hadoop2.7')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pyspark\n",
    "from pyspark import SparkContext,SparkConf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyspark.sql import SparkSession"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/sdmohant/spark-3.0.0-bin-hadoop2.7/python/pyspark/context.py:220: DeprecationWarning: Support for Python 2 and Python 3 prior to version 3.6 is deprecated as of Spark 3.0. See also the plan for dropping Python 2 support at https://spark.apache.org/news/plan-for-dropping-python-2-support.html.\n",
      "  DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "#spark = SparkSession.builder.appName(\"MyApp\").getOrCreate()\n",
    "#spark = SparkSession.builder.appName(\"MyName\").getOrCreate()\n",
    "conf = SparkConf().set(\"spark.cores.max\", \"16\") \\\n",
    "    .set(\"spark.driver.memory\", \"16g\") \\\n",
    "    .set(\"spark.executor.memory\", \"16g\") \\\n",
    "    .set(\"spark.executor.memory_overhead\", \"16g\") \\\n",
    "    .set(\"spark.driver.maxResultsSize\", \"0\") \\\n",
    "    .set(\"spark.sql.shuffle.partitions\",\"300\")\n",
    "\n",
    "sc = SparkContext(appName=\"sociologymodel\", conf=conf)\n",
    "spark = SparkSession(sc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyspark.sql.functions import isnan, count, col, when"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "sociologyFOSID = spark.read.options(header = True, inferSchema = True,delimiter=',').csv(\"FOSDomains.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "sociologyFOSID = sociologyFOSID.filter(sociologyFOSID[\"Domain\"]==\"Sociology\").collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "papers = spark.read.options(header = True, inferSchema = True,delimiter=',').csv(\"AllPapersFOS1.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyspark.sql.types import ArrayType, StructField, StructType, StringType, IntegerType,LongType"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "schema = StructType([\n",
    "    StructField('FOS_ID', LongType(), True),\n",
    "    StructField('Domain', StringType(), True)\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Convert list to RDD\n",
    "rdd = spark.sparkContext.parallelize(sociologyFOSID)\n",
    "\n",
    "# Create data frame\n",
    "fos = spark.createDataFrame(rdd,schema)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------+---------+\n",
      "|  FOS_ID|   Domain|\n",
      "+--------+---------+\n",
      "|24667770|Sociology|\n",
      "|36289849|Sociology|\n",
      "+--------+---------+\n",
      "only showing top 2 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "fos.show(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "papers_filtered = papers.join(fos,on=\"FOS_ID\",how=\"inner\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "7609703"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "papers_filtered.count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "papers_filtered = papers_filtered.withColumnRenamed(\"CitationCount\",\"label\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['FOS_ID',\n",
       " 'PaperID',\n",
       " 'Rank',\n",
       " 'PaperYear',\n",
       " 'Volume',\n",
       " 'Issue',\n",
       " 'ReferenceCount',\n",
       " 'label',\n",
       " 'EstimatedCount',\n",
       " 'PaperAge',\n",
       " 'IsJournal',\n",
       " 'min_FOSRank',\n",
       " 'max_FOSRank',\n",
       " 'max_FOSPaperCount',\n",
       " 'min_FOSPaperCount',\n",
       " 'min_FOSCitationCount',\n",
       " 'max_FOSCitationCount',\n",
       " 'PaperRefCnt',\n",
       " 'Domain']"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "papers_filtered.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "papers_filtered1 = papers_filtered.select([\"PaperID\",\"EstimatedCount\",\"Rank\",\"ReferenceCount\",\"PaperAge\",\\\n",
    "                        \"IsJournal\",\"min_FOSRank\",\"max_FOSRank\",\"max_FOSPaperCount\",\"min_FOSPaperCount\",\"min_FOSCitationCount\",\\\n",
    "                        \"max_FOSCitationCount\",\"PaperRefCnt\",\"label\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "(sample1,sample2) = papers_filtered1.randomSplit([0.09,0.5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1161261"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample1.count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import model_selection\n",
    "from sklearn.model_selection import KFold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------+\n",
      "|EstimatedCount|\n",
      "+--------------+\n",
      "|             0|\n",
      "|             0|\n",
      "|             0|\n",
      "|             0|\n",
      "+--------------+\n",
      "only showing top 4 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "papers_filtered1.select([\"EstimatedCount\"]).show(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "sample1_pd = sample1.toPandas()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_est = np.array(sample1_pd[\"EstimatedCount\"])\n",
    "y_act = np.array(sample1_pd[\"label\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Comparison with Microsoft Predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Mean Absolute Error of EstimatedCount:', 2.321830320660041)\n",
      "('Mean Squared Error of EstimatedCount:', 5698.641732564858)\n",
      "('Root Mean Squared of with EstimatedCount:', 75.48934847092573)\n"
     ]
    }
   ],
   "source": [
    "print('Mean Absolute Error of EstimatedCount:', metrics.mean_absolute_error(y_act, y_est))\n",
    "print('Mean Squared Error of EstimatedCount:', metrics.mean_squared_error(y_act, y_est))\n",
    "print('Root Mean Squared of with EstimatedCount:', np.sqrt(metrics.mean_squared_error(y_act, y_est)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "sample1_pd = sample1_pd.drop([\"PaperID\",\"EstimatedCount\"],axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1161261, 12)"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample1_pd.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = sample1_pd.iloc[:, 0:11].values\n",
    "y = sample1_pd.iloc[:, 11].values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import KFold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "kf = KFold(n_splits=5)\n",
    "kf.get_n_splits(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestRegressor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Mean Absolute Error:', 1.6566836384582513)\n",
      "('Mean Squared Error:', 226.19797413417703)\n",
      "('Root Mean Squared Error:', 15.039879458764855)\n",
      "('Mean Absolute Error:', 1.2008485417478654)\n",
      "('Mean Squared Error:', 115.72705828623891)\n",
      "('Root Mean Squared Error:', 10.757651150982676)\n",
      "('Mean Absolute Error:', 2.725717519631336)\n",
      "('Mean Squared Error:', 2339.1999550135297)\n",
      "('Root Mean Squared Error:', 48.365276335543975)\n",
      "('Mean Absolute Error:', 2.03822497899277)\n",
      "('Mean Squared Error:', 639.33629740551)\n",
      "('Root Mean Squared Error:', 25.285100304438384)\n",
      "('Mean Absolute Error:', 0.7495968761199996)\n",
      "('Mean Squared Error:', 570.5711203007362)\n",
      "('Root Mean Squared Error:', 23.886630576553408)\n"
     ]
    }
   ],
   "source": [
    "rmse_list = []\n",
    "for train_index, test_index in kf.split(X,y):\n",
    "    #print(\"TRAIN:\", train_index, \"TEST:\", test_index)\n",
    "    X_train, X_test = X[train_index], X[test_index]\n",
    "    y_train, y_test = y[train_index], y[test_index]\n",
    "    #print(\"X_train:\", len(X_train), \"X_test:\", len(X_test))\n",
    "    #print(\"y_train:\", len(y_train), \"y_test:\", len(y_test))\n",
    "    regressor = RandomForestRegressor(n_estimators=100, max_depth=30, random_state=0)\n",
    "    regressor.fit(X_train, y_train)\n",
    "    y_pred = regressor.predict(X_test)\n",
    "    rmse_list.append(np.sqrt(metrics.mean_squared_error(y_test, y_pred)))\n",
    "    print('Mean Absolute Error:', metrics.mean_absolute_error(y_test, y_pred))\n",
    "    print('Mean Squared Error:', metrics.mean_squared_error(y_test, y_pred))\n",
    "    print('Root Mean Squared Error:', np.sqrt(metrics.mean_squared_error(y_test, y_pred)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Mean RMSE is :', 24.66690756525666)\n"
     ]
    }
   ],
   "source": [
    "rmse_sum = 0\n",
    "for index in range(0,5):\n",
    "    rmse_sum = rmse_sum + rmse_list[index]\n",
    "mean_rmse = rmse_sum/5\n",
    "print(\"Mean RMSE is :\",mean_rmse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "featureslist=sample1_pd.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "importances = regressor.feature_importances_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Feature: 0, Score: 0.62676\n",
      "Feature: 1, Score: 0.01115\n",
      "Feature: 2, Score: 0.27903\n",
      "Feature: 3, Score: 0.00235\n",
      "Feature: 4, Score: 0.00842\n",
      "Feature: 5, Score: 0.01264\n",
      "Feature: 6, Score: 0.00600\n",
      "Feature: 7, Score: 0.01457\n",
      "Feature: 8, Score: 0.02874\n",
      "Feature: 9, Score: 0.00597\n",
      "Feature: 10, Score: 0.00437\n"
     ]
    }
   ],
   "source": [
    "for i,v in enumerate(importances):\n",
    "\tprint('Feature: %0d, Score: %.5f' % (i,v))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## XGBoost Regression model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "import xgboost as xgb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "kf = KFold(n_splits=5)\n",
    "kf.get_n_splits(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Mean Absolute Error of XGBoost Kfold:', 1.7361924149771961)\n",
      "('Mean Squared Error of XGBoost Kfold:', 321.7522005786695)\n",
      "('Root Mean Squared Error of XGBoost Kfold:', 17.937452455091538)\n",
      "('Mean Absolute Error of XGBoost Kfold:', 1.2551482307221948)\n",
      "('Mean Squared Error of XGBoost Kfold:', 132.21453035490288)\n",
      "('Root Mean Squared Error of XGBoost Kfold:', 11.498457738101353)\n",
      "('Mean Absolute Error of XGBoost Kfold:', 2.9192981926691175)\n",
      "('Mean Squared Error of XGBoost Kfold:', 2603.875964005259)\n",
      "('Root Mean Squared Error of XGBoost Kfold:', 51.02818793574057)\n",
      "('Mean Absolute Error of XGBoost Kfold:', 2.1252180171605866)\n",
      "('Mean Squared Error of XGBoost Kfold:', 627.6022456462018)\n",
      "('Root Mean Squared Error of XGBoost Kfold:', 25.051990851950308)\n",
      "('Mean Absolute Error of XGBoost Kfold:', 0.7476966204520299)\n",
      "('Mean Squared Error of XGBoost Kfold:', 401.322632815465)\n",
      "('Root Mean Squared Error of XGBoost Kfold:', 20.033038531772085)\n"
     ]
    }
   ],
   "source": [
    "rmse_list = []\n",
    "for train_index, test_index in kf.split(X,y):\n",
    "   # print(\"TRAIN:\", train_index, \"TEST:\", test_index)\n",
    "    X_train, X_test = X[train_index], X[test_index]\n",
    "    y_train, y_test = y[train_index], y[test_index]\n",
    "    regressor = xgb.XGBRegressor(\n",
    "                n_estimators=100,\n",
    "                reg_lambda=1,\n",
    "                gamma=0,\n",
    "                max_depth=40\n",
    "                )\n",
    "    regressor.fit(X_train, y_train)\n",
    "    y_pred = regressor.predict(X_test)\n",
    "    rmse_list.append(np.sqrt(metrics.mean_squared_error(y_test, y_pred)))\n",
    "    print('Mean Absolute Error of XGBoost Kfold:', metrics.mean_absolute_error(y_test, y_pred))\n",
    "    print('Mean Squared Error of XGBoost Kfold:', metrics.mean_squared_error(y_test, y_pred))\n",
    "    print('Root Mean Squared Error of XGBoost Kfold:', np.sqrt(metrics.mean_squared_error(y_test, y_pred)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Mean RMSE is :', 25.10982550253117)\n"
     ]
    }
   ],
   "source": [
    "rmse_sum = 0\n",
    "for index in range(0,5):\n",
    "    rmse_sum = rmse_sum + rmse_list[index]\n",
    "mean_rmse = rmse_sum/5\n",
    "print(\"Mean RMSE is :\",mean_rmse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "importances = regressor.feature_importances_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fbbb0e1cfd0>"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "(pd.Series(importances, index=featureslist[0:11]).plot(kind='barh')) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Feature: 0, Score: 0.35229\n",
      "Feature: 1, Score: 0.06545\n",
      "Feature: 2, Score: 0.06232\n",
      "Feature: 3, Score: 0.02309\n",
      "Feature: 4, Score: 0.09034\n",
      "Feature: 5, Score: 0.10763\n",
      "Feature: 6, Score: 0.03497\n",
      "Feature: 7, Score: 0.12514\n",
      "Feature: 8, Score: 0.09638\n",
      "Feature: 9, Score: 0.02461\n",
      "Feature: 10, Score: 0.01777\n"
     ]
    }
   ],
   "source": [
    "for i,v in enumerate(importances):\n",
    "\tprint('Feature: %0d, Score: %.5f' % (i,v))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.15rc1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}